Internet Explorer 11 is not supported

For optimal browsing, we recommend Chrome, Firefox or Safari browsers.

Cities Hope AI Camera Enforcement Can Improve Bus Service

Philadelphia, New York and Washington, D.C., are using bus-mounted cameras with AI technology to better enforce parking violations, hoping to clear transit lanes of vehicles and make public transit faster and safer.

IMG_3239 (002).jpg
A parked vehicle blocks a bus lane in downtown Philadelphia. (Jared Brey/Governing)
In Brief:
  • Philadelphia is planning to launch automated enforcement of parking violations at bus stops and in bus lanes using bus-mounted cameras.

  • Washington, D.C., launched a similar program last month, and New York has had one in place since 2019.

  • Automated enforcement improves compliance and speeds up bus service, but needs to be built on public trust, proponents say.

  • Buses crawl through Center City Philadelphia. On two of Southeastern Pennsylvania Transportation Authority’s (SEPTA) busiest routes, cutting across the heart of town, vehicles don’t even crack 9 miles per hour on average. Idling in downtown congestion — running up staff time and burning fuel while fare-paying riders wait at the curb — costs SEPTA an average of $15 million annually, according to the agency. And it’s not just because there’s too many cars in general, but because too many drivers park illegally at bus stops and in bus lanes.

    Last spring, when SEPTA ran a pilot program to test the efficacy of bus-mounted cameras that automatically record instances of illegal parking in transit-only spaces, it found “an alarming amount of violations,” according to a report: 36,392 total obstructions in just two months, on just two routes, using cameras on just seven individual vehicles. Alarming but not surprising, says Matt Zapson, a planning project manager at SEPTA.

    “We hear it from our operators and from our riders every day,” Zapson says. “It’s a challenge that makes the bus less accessible, makes it less safe at bus stops, and it’s one of the reasons why our buses are so slow in Center City and can be unreliable.”

    SEPTA used the results of the study to push for authorization to use automated enforcement in Center City permanently. When it rolls out the new program sometime next year, it’ll join a growing list of cities turning to automated enforcement of traffic violations in spaces designated for transit. The programs, powered by sophisticated artificial intelligence technology, throw the shortcomings of existing enforcement methods into sharp relief. And cities like Philadelphia hope they can make bus service faster, more reliable and safer.

    Accuracy and Consistency


    SEPTA’s pilot program, which involved recording violations but not issuing citations, was powered by Hayden AI, a San Francisco-based “mobile perception platform” that works with other cities, including New York and Washington, D.C., to provide automated enforcement of certain traffic violations. The company works with cities and transit agencies to map roadways, which are then “annotated” with bus stops and transit-only lanes, says Charles Territo, Hayden AI’s chief growth officer.

    It then installs outward-facing cameras on buses. When those cameras observe obstructions in transit-only lanes, they begin recording short segments of video. They send a digital package of evidence, including license plate information, to an enforcement agency, which can then quickly send citations to drivers.

    The automation makes enforcement exponentially more accurate and consistent than existing methods, Territo says.

    “We are using AI and not people to identify the objects, to map the roads, to almost digitize the city. And it allows us to very efficiently and very easily make the types of determinations we need to make when it comes to the violations,” he says.

    New York began using automated bus lane enforcement in 2019. In combination with other operational improvements, the method helped increase bus speeds between 15 and 30 percent on enforced corridors, according to the New York City Department of Transportation. With fewer obstructions, buses are less likely to merge into other vehicular lanes, which helps reduce collisions substantially as well, New York has found. The vast majority of people who receive a ticket never get a second one. New York Gov. Kathy Hochul signed a law permitting expansion of the program earlier this year.

    Washington, D.C., rolled out an automated enforcement program on Washington Metropolitan Area Transit Authority buses last month, after a period of sending warnings to drivers caught in violation of traffic laws. Hayden AI cameras were mounted on 140 buses running 31 routes. In California, the Alameda-Contra Costa Transit District is working to roll out a program next year as well, as is Los Angeles Metro.

    Building Public Trust


    At a practical level, automated enforcement is vastly better than the existing system, which involves inconsistent, expensive and often conflict-ridden interactions between enforcement officers and drivers, says Lindiwe Rennert, a senior research associate in the Metropolitan Housing and Communities Policy Center at the Urban Institute. When considered against the backdrop of rising traffic deaths, it’s clear that money spent on enforcement now is “very close to dollars down the drain.”

    “The current practice is so bad — provenly bad, measurably bad, demonstrably bad,” Rennert says. “It’s ineffective and costs us a lot.”

    While automating enforcement is “a step in the right direction,” she says, there are important elements of public trust-building and equity that cities need to consider as they develop programs, says Rennert, who has conducted focus groups on related issues with Black civic leaders. People need to feel confident that the programs are about making streets safer, not about generating general revenue, she says. Fines should be “right-sized” to avoid overburdening low-income people, who may be more likely to receive citations based on the locations of enforced spaces. Ideally, the programs should be paired with — and perhaps help to fund — efforts to build safer street infrastructure, like curb bump-outs at bus stops. In a perfect world, that would help make the camera programs obsolete, Rennert says.

    The biggest obstacle to the expansion of automated enforcement, Rennert says, is “public distaste.” But it’s not the only one. Cities also operate under a patchwork of state laws, similar to those governing automated enforcement of speeding and red-light violations. Cooperation between local transit agencies, departments of transportation and parking enforcement agencies can also be a challenge.

    In Philadelphia, the much-maligned Philadelphia Parking Authority (PPA) will act as the “system administrator” for the automated bus lane enforcement program. The PPA is expected to select a vendor next year. Zapson says SEPTA has seen speed improvements from adding bus-only lanes and from painting the pavement red. Better enforcement won’t eliminate parking violations altogether, but it should deter enough people to speed up service on the city’s busiest routes, he says.

    “We’re hopeful and we believe that providing that kind of service — safe, accessible, fast, reliable — will draw more people to the system,” Zapson says.
    Jared Brey is a senior staff writer for Governing. He can be found on Twitter at @jaredbrey.
    Special Projects